1
Thesis
Name: Andreea Panțuru
Student number: 10983511
Specialisation: Economics and Finance
Field: Development economics (Macroeconomics)
Number of credits: 12 EC
Title: The correlation between China’s One-Child Policy (1980) and
fertility
2
Statement of originality
This document is written by Student Andreea Panțuru who declares to take full responsibility
for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources
other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion
3 Table of Contents
Abstract ... 4
1. Introduction ... 4
2. Literature review ... 6
i. The Malthusian model (end 18th century) ... 6
ii. The Demographic Transition Model (mid-20th century) ... 7
iii. Underlying reasons for the Demographic Transition ... 9
iv. Situation in China ... 13
3. Methodology ... 15
i. Data ... 15
ii. Research method ... 18
4. Results ... 19
Discussion ... 21
Conclusion ... 24
Reference list ... 26
4 Abstract
This paper investigates the effect of China’s One-Child policy on de downward trend of fertility levels that followed after its implementation. It aims to give a better understanding of the underlying reason and possible solutions for the current phenomenon of negative population growth in which fertility rates have reached beyond-replacement levels. The results indicate that the One-Child policy had a negative, but insignificant effect on fertility levels and that it is socio-economic factors that affect fertility significantly. This suggests that to fix nowadays issue of a declining population, the focus should lie on socio-economic factors instead of government policies only.
1. Introduction
Over the past decades, the total population accounted a growth of four billion people (Bongaarts, 2009). And as a response, governments of developed countries tried to control the population through policies. (Bongaarts, 1994) Over time a new phenomenon of declining fertility rates and population emerged (Doepke, 2004). But how did this happen? Moreover, how can we influence population size and enhance birth rates? To provide an answer, the following question is investigated in this research: Can government policies significantly influence fertility rates? By learning more about the effect of family planning on the downward trend of fertility, we distinguish both the cause and a possible solution for the issue of a graying population. To do so, this paper focuses on one of the largest population control policy experiment implemented, China’s One Child Policy, and its effect on fertility rates. The aim of this paper is to provide insight in the degree to which and how governments can help solve the worldwide trend of prolonged negative population growth the world is now facing. In other words, if previous population control policies are indeed found to be fundamental in decreasing birth rates, it can be argued that they can also be used to reverse this trend.
The demographic transition depicts the many demographic changes the world has experienced in previous years. New challenges emerged after the high population growth made place for the phenomenon of population graying (Bongaarts, 2009; Bloom et al., 2011). A good explanation for the population ‘becoming older’ results from the combination of the downward fertility trend combined with the initial increase in number of people. Moreover, countries such as Germany and Japan have recently reached a point in which death rates are higher than birth rates. This, over the long term, is not sustainable. Among other, concerns regarding a country’s
5
economy and for instance the operation of pension and health care systems arise (Bloom et al., 2011). An example in attempting to solve this issue are parental friendly laws such as the Parenteral allowance and Parental Leave act introduced by Germany in 2007 (Schultz, 2016). Past policies can help form accurate expectations regarding the potential of current policies in controlling the population and addressing its decline.
Great examples of governments trying to control population can be observed in China and India. There, family planning became a central theme around 1970 as further population growth was believed to hinder the overall progress of a country (Banerjee & Duflo, 2011). This is why, from mid-1975 until early 1977, India entered the Emergency, an era that enforced a sterilization policy. This policy gave each state the right to impose sterilization laws and forced them to meet certain sterilization quotas. This resulted in almost a quarter of the Indian couples becoming infertile by 1976 (Banerjee & Duflo, 2011). Severe measures were taken to accomplish this. For example, it was very common for teachers to pressure parents into sterilization by threatening to deny their children’s school enrollment or for railway inspectors to give either huge fines or the option of sterilization to people without a ticket (Banerjee & Duflo, 2011).
Looking back, as drastic as India’s policy seems as fast did its effect on population disappear. One of the reasons for this is that it only spread over little time (Banerjee & Duflo, 2011). Thus, to address this issue, a policy maintained over a longer period should be assessed. China’s One-Child policy enacted in 1980 fits the description. By being another renown example of strictly enforced population control measure and one of the largest and longest experiment in fertility policy administered in human history, the One-Child policy can help shed light into the research question. As the name states, its aim was to reduce couples to having only one child. This was enforced through high fees and severe measures such as forced late-term abortions. While often unreported, dreadful stories of woman pregnant in their eighth month forced to abort were a phenomenon all over China (Wen, 2014).
So, as opposed to the case in India, China’s initial family planning policy has been valid up until 2015 and therefore gives the opportunity to investigate its significance on fertility over a longer term (Cai & Wang, 2006). Recently, the government changed it to a ‘Two-Children Policy’, indicating the current concerns regarding a negative population growth observed all over the world. To get more insight in a possible solution for this issue, this paper will research the relationship between fertility and government policies through a panel data analysis applying the Difference-in-Difference method. It constitutes a robustness check on previous
6
literature that investigates the effect of the One-Child policy on fertility. It will be done over the year-span of 1975 till 2016 and applied on a sample of twelve Asian countries. In the second section of this paper the theoretical framework is discussed. The methodology can be found in section three, where the data and research method are outlined. Section four provides the results and is followed by the discussion in section five. Lastly, section six concludes this paper’s research.
2. Literature review
The total fertility rate (TFR) refers to the expected number of births per average woman. To calculate the TRF, age-specific fertility rates over intervals of five years are added up. The according equation can be found below (OECD, 2014):
TFR = 5 * ∑ (age-specific fertility rates) = 5 * ( 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑖𝑟𝑡ℎ𝑠 𝑡𝑜 𝑤𝑜𝑚𝑒𝑛 𝑎𝑔𝑒𝑑 15−19
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑤𝑜𝑚𝑒𝑛 𝑎𝑔𝑒𝑑 15−19 + ⋯ + 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑖𝑟𝑡ℎ𝑠 𝑡𝑜 𝑤𝑜𝑚𝑒𝑛 𝑎𝑔𝑒𝑑 45−49
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑤𝑜𝑚𝑒𝑛 𝑎𝑔𝑒𝑑 45−49 )
The replacement level of birth rates indicates the total amount of births a woman should for the population to be stable and sustainable. This level amounts a TFR of 2.1 children per woman (OECD, 2014). Bongaarts (2002) claims that since the 1950s, the developed world has experienced a 44 percent drop of total fertility rates. More exact, over a period of only 40 years, this rate went from 2.8 births to 1.57 births per woman. This downward trend in fertility observed worldwide, placed the birth rates in many developed countries beyond-replacement level (OEDC, 2014).
The literature review follows a chronological timeline regarding models. First, the Malthusian model, also known as the first theory of population, is introduced. Then, the model of Demographic Transition and its five stages is discussed. Subsequently, the reasons underlying this phenomenon of transition are explained and supported with economic theories on fertility. Lastly, the focus lies on China and different views on the effect of its One-Child Policy.
i. The Malthusian model (end 18th century)
At the end of the eighteenth century, Thomas Robert Malthus introduces a theory on population. His main principle was the following: “Population, when unchecked, increases in a geometrical ratio. Subsistence increases only in an arithmetical ratio” (Malthus, 1987). Thus, Malthus
7
disequilibrium. As food is essential for survival, shortages pose big concerns. He addresses this issue by introducing positive and preventive checks. Those checks attempt to deal with this unbalance by slowing down population growth (Malthus, 1987).
Positive checks are of a natural origin and include all factors that shorten a person’s life. Think of for instance wars, shortages of food, poverty, diseases etc. Preventive checks on the other hand are created by man-kind. They include measures such as late marriage, self-restraint and family planning policies (Levin, 1938). These checks make the Malthusian model relevant for this study as it looks at their capability of affecting population. Moreover, the checks are comparable to the factors researched in this paper. Namely the One-Child policy against natural socio-economic factors is equivalent to respectively preventive against positive checks.
Furthermore, Malthus claims that a country has fixed resources. From this, the assumption can be drawn that rich countries have a low population growth and vice versa (Galor & Weil, 2000). However, this theory has its limitations. While population grew substantially, most of the countries also became richer. This contradiction with Malthus his assumptions can be blamed on unanticipated technological progress. Also the occurring downward trend of fertility plays a big role in people becoming richer which on the other hand, is line with Malthus his theory. Still, taking all developments into account, Banerjee and Duflo (2011) argue that it may be more meticulous to say that for instance it is poverty that triggers high birth rates, and not the other way around.
ii. The Demographic Transition Model (mid-20th century)
The Demographic Transition Model (DTM) was introduced mid-20th century and demonstrates the trend of population, triggered by changes in fertility and mortality levels (Hilgeman & Butts, 2009). Based on an analysis by Warren Thompson (1929), this model contains five stages that categorize and depict the changes in birth and death rates occurring mainly after the Industrial Revolution. As discussed by Doepke (2004) and shown below, the fertility-rate decreases in most developed countries and eventually reaches a beyond-replacement level.
Graphical support is provided for the five stages of the model. Examples of different countries are shown for each stage. In the graphs, the replacement-level of 2.1 births per woman is indicated by a reference-line. The green scatterplot represents the total population growth in percentages. Furthermore, the mortality and fertility rates are depicted by the red and blue scatterplots. Note that those rates are originally measured in different ways, which is why mortality rates seem higher than fertility rates overall. In practice however, this is surely not the
8
case. But as the focus lies on the trend, and not actual rates, it is beyond the scope of the graphs to address and correct for this issue.
Stage 1, known as the Pre-transition phase, is characterized by both high birth and
mortality rates resulting in a balanced overall population growth. This era is pre-industry, meaning that no graphical material can be shown as today all countries are industrialized.
Stage 2, the Early-transition, death rates start falling while birth rates remain stable.
According to Hopfenberg (2014), better hygiene and medical developments made such movement possible. As a result, this has a ‘natural increase’ in population and indicates a rapid growth. Countries within Sub-Saharan Africa, Afghanistan and Guatemala are currently in stage 2 of the DTM.
Stage 3 is called the Late-transition, where population growth slows down as birth rates
start to decline. An uprising industrialized society where children become an extra cost can explain this trend (Hopfenberg, 2014). Countries such as Indonesia, India and Vietnam are currently in stage 3 of the DTM. Note that for India, the purple reference line indicates the sterilization policy introduced in 1977.
As countries enter stage 4, the Post-transition phase, low birth rates and mortality rates are observed. Population stops growing and therefore is stable, risking starting to decline. Currently, many developed countries are in this stage such as China and Korea, Rep. Also, they both face a high risk of entering stage 5 (Hopfenberg, 2014). Additionally, as this research
9
focuses on the One child policy, China’s graph shows the policy by a purple vertical reference line in 1980.
Stage 5 is the last phase in the demographic transition model where death rates are
higher than birth rates and countries lose their population. Countries like Germany, Japan and Ukraine are currently in stage 5 (Hopfenberg, 2014).
So, after looking into the Demographic Transition Model, we can draw the conclusion that the decline in population growth is a current, world-wide phenomenon. This model supports the aim of the paper by pointing out the issue of population aging. On that account, the DTM emphasizes the relevance of finding a solution to this issue.
iii. Underlying reasons for the Demographic Transition
This section investigates the origin and reasons of the Demographic Transition, focusing on what drove the decline in fertility rates. Research from the OECD (2014) mentions various reasons. These can be divided in two categories, respectively ‘internal factors’ and ‘external factors’. Internal factors affecting fertility are from a household’s perspective. Think of income changes, higher female education and more women joining the workforce. External factors on the other hand, lie beyond the household and refer to indicators such as government policies and urbanization.
10 a) Internal factors
The slogan ‘A small family is a happy family’ portrays the somewhat old-fashioned analysis on fertility introduced by Gary Becker in 1960 (Banerjee & Duflo, 2011). In his model, children as treated like durable consumer goods. His approach has two main characteristics. First, to add economic value to the model, Becker assumes that people have set preferences and excludes differences in taste. He infers that fertility movements should mainly be derived from changes in income. Second, the ‘quantity-versus-quality tradeoff’ is central and addresses the relation between income and fertility. The essence of this trade-off is that as households get richer, they will rather invest more in less children than have more children. Moreover, Becker considers the following factors to affect fertility levels: income, the cost of children and knowledge regarding birth control (Doepke, 2014).
Aside from technological progress, higher incomes and demand for human capital emerged due to the Industrial Revolution. This development had two main effects on the population growth (Galor & Weil, 2000). First, in line with the ‘quantity-quality trade-off’, a higher income provided more resources that were directed towards investing in children (Oded, 2012). Second, this reallocation of resources focused on the quality of children and suggested a reduction in fertility (Galor & Weil, 2000). Furthermore, an occurrent pattern regarding family formation is emerging in the developed world: people are more likely to establish their professional life before founding a family and prefer a family size of less than 2 children (Hilgeman & Butts, 2009). Thus, all these factors infer that higher income levels contribute to a decline in birth rates.
Other factors influencing fertility levels are education and joining the workforce. According to the OECD (2011), education and labor participation amongst females has risen substantially over the last 30 years. First, technological progress stemming from the Industrial Revolution created more jobs. Furthermore, as the quality of children became more important, the average level of education increased accordingly (Oded, 2012). So, as better opportunities arose, both female education and labor participation increased. But since both children and jobs are time-consuming, due to time-constraints women will opt for less child-births (Hilgeman & Butts, 2009).
Moreover, the degree of education among women is highly correlated to the number of births they favor (OECD, 2011). For instance, it is more likely that women with tertiary education remain childless when compared to women with secondary education. Complications
11
such as a lack of childcare services and higher overall living prices combined with more women joining the workforce make a stable job necessary and having children less appealing (OECD, 2004). Additionally, when receiving a better education, women acquire more knowledge on birth control, further contributing to a decline in fertility (Becker, 1960; Easterlin, 1978).
b) External reasons
According to Becker (1960), the cost of children is influence by child mortality, government policies and urbanization. While the decreasing trends of child mortality decreased the costs of children, urbanization had the opposite effect. Government policies could go either way depending on their aim (Becker, 1960; Oded, 2012). Child mortality is argued to be a possible factor in influencing the demographic transition (Oded, 2012). Lower child mortality indicates that less costs are attached to creating a child of high quality, since less births are necessary for ensuring its survival (Becker 1960).
Regarding urbanization, this seems to have a direct negative impact on fertility as education and living arrangements are more expensive in urban areas (Becker, 1960). Another argument is that urban households have several advantages over rural households. Those advantages refer to most factors that generally have a direct effect on fertility such as education and labor participation, equal opportunities and access to multiple types of services (Martine, Alves & Cavenaghi, 2013). Additionally, Cai (2010) adds that urbanization can interrupt or delays the start of a family and thereby affect fertility.
Through policy implementation, governments try to control population. And in the past, as countries were in the beginning stages of the demographic transition, most policies were focused on increasing the cost of children and therefore reducing fertility levels (Bongaarts, 1994; Becker 1960). Meanwhile, with an eye on the current threat of below-replacement fertility levels, governments attempt to boost birth rates by implementing reversed, fertility-enhancing policies such as maternal and paternal leave, flexible working hours, free health benefits and childcare possibilities. Also, they aim to prevent that children become a burden for parents making it easier for them to pursue their professional regardless of the number of children they opt to have (Hilgeman & Butts, 2009).
c) Economic analysis of fertility
Adding onto Gary Becker’s view, Richard Easterlin (1978) includes potential social factors such as cultural differences, religion and relative income between generations in his model on
12
fertility (Doepke, 2014). He argues that an approach focused on demand towards fertility behavior such as introduced by Becker may not hold in many scenarios. Whereas his model, that considers output shifts alone, separately form any demand conditions, is more adequate to deal with real world cases (Doepke, 2014). Therefore, the main focus when providing theoretical support for the internal and external factors mentioned above is put on Richard Easterlin’s model.
This model introduces concepts such as natural fertility and accounts for alternative hypotheses that are most likely to be expressed by non-economists and sociologists. However, even with a broader empirical base than prior theory on fertility, it still does not manage to address all factors affecting birth rates. Nevertheless, Easterlin argues that fertility indicators work through one or more of the following measures: the demand for children, the potential output of children and all costs related to fertility regulation (Easterlin, 1978).
First of all, household’s income is central in this model as it influences the demand for children. The demand for children refers to the number of births a household wishes when fertility regulation is without charge. Then, the potential output of children depends on the baby’s chance of surviving till adulthood as well as the overall natural fertility. It is depicted by the number of births arising if fertility is not limited on purpose. Lastly, the costs attached to fertility regulation refer to the time and money needed to learn about the usage of birth control techniques and accounts for the way people view these (Easterlin, 1978).
In short, Easterlin (1978) mentions two cases. Excess demand, when the potential output of children is lower than the demand and households try to increase fertility through for example adoption. In case of excess supply, there is a significant incentive to make use of fertility control as households have more births than preferred. The degree of the actual births being regulated is shown by the difference between potential output and actual number of children. Whereas the unwanted births are found by comparing the actual number of children with the desired birth number.
Finally, it is important to note that practices limiting fertility are highly relevant in fertility analyses based on demand. (Easterlin, 1975) Such practices could be self-restraining, but also family planning imposed by governments. For the latter, Easterlin (1975) emphasizes the importance of the reason behind using such practice. These reasons should be strong enough as they are the main driver behind the success of a family planning program. A strong enough reason gives an incentive to people to demand family planning programs, increasing their
13
overall efficacy and success rate. All in all, in line with this reasoning, government policies on their own are limited in affecting population without people willing and having a reason to make the changes they impose.
iv. Situation in China a) The One-Child Policy
When determining the development of population, the level of fertility is the main demographic factor that needs to be accounted for (Jiang, 2010). Looking at the Chinese level of fertility starting from 1979 onwards, a downward trend is observed. In 1979, the birth rate was as high as 2.9 children per woman which decreased to 1.7 by 2004 and to only 1.4 by 2010 (Cai & Wang, 2006). As mentioned previously, socio-economic factors as well as governments can influence fertility rates. Regarding government policies specifically, China differs from the rest of the world by its One-Child Policy enacted in 1980 that may have significantly contributed to the downward fertility trend on top of the socio-economic aspects.
In 1979, China’s family planning policy was implemented by the government. Its aim was to control the population growth the country was facing by allowing each woman only one child and imposing fines for every additional child (Cai, 2010). So, heavy penalties proportional to the salary of a household were attached to exceeding the enforced quotas (Li, Zhang & Zhu, 2015). At worst these could constitute 70% of a households’ salary, being permanently exempt from bonuses or promotions and making the ‘children above the quota’ illicit from public urban schools. It is important to note that this policy showed substantial heterogeneity, which can be recognized from stricter rules in urban areas compared to rural ones regarding both quotas and fines (Li, Zhang & Zhu, 2015). It can be argued that such heterogeneity indicates the difficulty of imposing a strict type of family planning control on a big group of people and therefore weakens the argument claiming that government policies can control the population as a whole.
b) Different views on the significance of the One-Child policy on fertility rates
Through the support of bureaucracy focusing on the control and enforcement of rules, the Chinese family planning policy discussed above managed to spread throughout the country in both urban and rural regions (Cai, 2010). According to the National Population and Family Planning Commission (NPFPC) of China in 2007, the One-Child policy was a huge success that managed to avert 400 million additional births. They add that indirectly, the policy led to China’s economic growth and helped against the global warming problem (Cai, 2010). Many researchers agree with this view and claim that the One-child policy significantly affected and
14
lastingly changed the demographic trend of fertility in China. An example would be the study of Yu (2008). He analyses the demographic development in China and concludes that the One-child policy had a fundamental effect on total birth rates. Also Settles and Sheng (2008) stress that its impact and degree of success should not be belittled. According to them, this policy slowed the growth of the population by 5.7% between 1979 and 2005, placing the fertility rate beyond the replacement level of 2.1 children per woman.
However, contrary to the NPFPC (2007), scientific research argues that the main reasons for the downward trend in China’s birth rates cannot not be attributed to a family planning policy only. Those are highly complex and should include and focus on developments within the society and economy. It is argued that mainly those factors played a big role in the rapid decline of fertility (Jiang, 2010). Garcia (2010) conducts an empirical research and, in line with the above, he documents that the One-Child Policy had little impact on the decrease in observed fertility after 1980. He executes his research as follow. To see whether socio-economic factors were the main driver behind the lower fertility rather than the outcome of the family planning policy, he secludes the policy from other main socio-economic factors affecting fertility. The factors Garcia (2010) chooses as independent variables affecting fertility are the following: a higher income and thus opportunity costs per child, higher education among females and the overall living of households becoming more expensive.
These observations are approved by many other scientists (e.g. Cai, 2010; Li, Zhang & Zhu, 2015). Just like Garcia, they find evidence that the decrease in fertility is highly driven by economic and social factors, more so than by governmental policies. Cai (2010) conducts an empirical research where he compares two Chinese provinces with each other: Jiangsu and Zhejiang. As hypothesis he claims that by having a more lenient policy, Zhejiang is expected to have higher birth rates. However, this hypothesis does not hold as fertility levels are found to be similar between the two regions. Four main socio-economic factors influencing fertility are included in his analysis, namely: GDP per capita, globalization (FDI per capita), women’s education level and changes in urbanization. So, his study suggests that the drive behind China’s below replacement fertility may have been affected and possibly fastened by the One-Child policy. But that it is socio-economic developments that play the main role in the development of fertility in China, as they also do for other countries and societies. (Cai, 2010) The fact that neighbor countries without such policy also experienced a decrease in fertility makes an important argument behind the insignificance of China’s policy on fertility.
15
So, to provide insight in the possible efficacy of present policies aimed at enhancing fertility, the research narrows down to China’s One-Child policy. As seen from above, even the effect of one of the largest and longest experiment in family planning is still unclear. The main difficulty in analyzing population control policies is that many other factors also play a big role and result in population changes. Then, questioning which particular factor is the main driver behind changes in fertility becomes central. Is China’s government right about its policy being responsible for the downward fertility trend or is there more to the story?
3. Methodology
After taking all socio-economic factors discussed in the literature review and used in previous empirical research into consideration, the indicators of fertility are selected based on their degree of relevance for Asia. Those indicators refer to the independent variables, with fertility being the dependent variable. Including China, twelve Asian countries are part of the sample. Furthermore, the time-span of 1975-2016 is analyzed. Below, the ideal regression is formulated together with the main hypothesis. Important to note is that β1 is the variable of interest as it
refers to China’s One-Child policy.
Fertilityit = β0 + β1 govt policiesit + β2 GDP per capitait + β3 female school enrollment, primary
+ β4 female labor force participation rate it + β5 urbanizationit + α1 infant mortality rateit + α2
contraceptive availabilityit + α3 social systemsit + εit
Where t stands for time and i stands for the country variable.
Null Hypothesis: β1 = 0
Alternative Hypothesis: β1 < 0 i. Data
First, the choice of independent variables is elaborated on. Then, an explanation is provided for the variables that need to be controlled for. Lastly, the method behind the research is discussed.
a) Independent Variables (β)
In this paper, the aim is to include the following indicators: GDP per capita (constant 2010 US$), School enrollment, primary, female (% gross), Labor force participation rate, female (% of female population ages 15+) (modeled ILO estimate) and Urbanization. As argued in the literature review, these are the main socio-economic factors influencing fertility rates and thus
16
also population growth. Their effect will be compared with the effect of the main independent variable of interest: Government Policy.
Government policy (Dummy): This independent variable indicates China’s One-Child
policy implemented in 1980. It is a dummy variable, meaning that for country ‘China’, the coefficient is 1 for the period of 1980-2016. As this dummy only applies on China’s policy, in all other cases the coefficient is 0, including previously discussed countries with other family planning policies such as India.
GDP per capita (constant 2010 US$) data was retrieved from the World Bank. GDP per
capita stands for the gross domestic product per person where inflation is not taken into consideration. As inflation is different among countries and influences GDP, by adjusting for its effects and taking it out of the equation the true growth is measured, making it easier to compare the actual household income between the sample of countries.
In line with the Malthusian model, many economists and academics have observed a strong negative relationship between GDP per capita and fertility rates. GDP is a good indicator for a country being either poor or rich and richer countries are believed to have lower population growth and vice versa (Banerjee and Dufly, 2011). This strong negative relationship can be explained by the Quantity-Quality model of Gary Becker as well as by the fact that low GDP per capita infers a lower cost of children as households’ time is cheaper.
School enrollment, primary, female (% gross) data was retrieved from the World Bank.
Female education has a direct and negative effect on fertility. Women receiving better education have a lower birth rate (Becker, 1960). Not only are educated women better capable of supporting themselves, but they also acquire more bargaining power regarding the formation and size of the family. Furthermore, they have more information on contraceptives and the means to form an opinion on their desired number of births (Banerjee and Duflo, 2011).
It should be accounted for the fact that the chosen indicator only looks at enrollment in primary schools and therefore may limit the results. However, while not optimal as a measure for the degree of education, it nevertheless provides the number of girls encouraged to start and continue an education, which positively affects the probability of women getting higher levels of education. Thus, due to lack of data, this indicator was chosen and is the best in indicating the education among females.
17 Labor force participation rate, female (% of female population ages 15+) (modeled ILO estimate) data was retrieved from the World Bank. There is a negative correlation between this
indicator and fertility rates as both children and having a job are very time-consuming. Since women are often responsible for taking care of the children, having a job results in a trade-off and contributes to less child-births (Hilgeman and Butts, 2009). One limitation regarding the use of this variable in the regression is no data exists prior to the 1990’s.
Just like educational attainment, labor force participation among females has considerably increased over the last 30 years (OECD, 2011). China, as well as many of the countries included in the sample have experienced huge economic growth during the analyzed timespan and as this happened more jobs and better opportunities became available. This led to higher educational levels among women and the provision of better workforce opportunities and resulted in a higher labor participation (Hilgeman & Butts, 2009).
Urbanization data was retrieved from the World Bank. This refers to the process of the
population moving to urban areas and indicates the percentage of people living in towns. It is almost in all cases so that urban fertility is lower than rural fertility, which is why urbanization has been acknowledged by analysts to be one of the main factors in fertility decline. According to Cai (2010), most countries selected in the sample have known a rapid urbanization during the timespan analyzed, making effect this had on fertility relevant. Additionally, also Becker (1960) agrees on the negative relation between fertility and urbanization. He states that moving from rural to urban settings raises the average cost of children which negatively affects the birth rate. After all, growing up in a farm is cheaper than in a town.
b) Control variables (α)
The following variables should be controlled for as they may affect the final results regarding the degree to which the variable of interest, government policies, affects the dependent variable compared to the socio-economic variables.
Infant mortality: When infant mortality is high, more births are needed to achieve the
favored number of children. Hence, lower infant mortality can prevent additional births. (Banerjee & Duflo, 2011) This effect should be considering for, as it can bias the results.
Welfare systems/ Social security net: Another factor biasing birth rates upwards is an
underdeveloped social system. In case of little or no social security, parents count on their children to take care of them when at old age, and thus aim to have a higher number of children. (Banerjee & Duflo, 2011)
18 Contraceptive availability: Evidence found by Banerjee and Dufly (2011) suggests that
the access to contraceptives alone has little direct effect on fertility levels. However, its possible effect, even when indirect, should be controlled for.
ii. Research method
This section elaborates on the steps taken to derive the main results of this research. Moreover, it goes in depth on the method used and analyses carried out.
a) Analysis of original regression
Unfortunately, the control variables ‘Contraceptive Prevalence’ and ‘Adolescent birth rates’ together with the independent variable ‘Female labor force participation rate’ listed in section three, cannot be included in the final regression. Due to lack of data for the years 1975-1980, their effect on fertility cannot be shown, as input is necessary to conduct a regression and this is simply not available. Therefore, an adjusted, more limited regression model emerges:
Fertilityit = β0 + β1 govt policiesit + β2 GDP per capitait + β3 female school enrollment, primary
+ β4 urbanizationit + α1 infant mortality rateit + εit
b) Fixed or Random effects
Since the regression is conducted on a panel data base, we must choose between using either fixed or random effects. To do so, the Hausman test is executed, where a highly significant result implies it is preferred to use fixed effects. After running the Hausman test on the adjusted regression, we get a result of p.=0.1681 which is insignificant at a 5% alpha level. This means that random effects are favored for this model. While fixed effects deal with omitted variable bias and can be used for a higher amount of countries, it does not consider the variation between those countries. Given our sample,
the variation between countries is crucial in determining how fertility is affected by the indicators and therefore, just like the Hausman test indicates, random effects should be used.
However, to properly assess the impact of the One-Child
19
policy implemented in China in 1980, the Difference in Difference (DID) quasi-experimental method will be applied, relying on the random effects of the panel data. DID is a technique used when studying the effect of a ‘treatment’ on a ‘treated group’ compared to the ‘control group’ by looking at the difference between those two. In this case, the treatment is the One-Child policy (1980), the treated group is China and the control group is a sample of eleven, similar Asian countries. DID excludes the effect of the One-Child policy on fertility by looking at the average change in fertility rates in China, compared to the average change in fertility rates for the rest of the countries in our sample that were not subjected to this policy. The graph above supports this explanation. In this way, the DID-estimator examines the effect of the treatment, here referring to the One-Child policy and tells whether it had a significant contribution on fertility.
4. Results
The adjusted, more limited regression stated in the second part of Section 3 is run, applying the Difference in Difference method and relying on random effects. This is done over a time span of 43 years: 1975-2016. Table 1 shows the main results of this analysis. The original linear regression on which the DID-method is applied can be found in (1), followed by a normalized version in (2). Limitations to these results are further discussed in section 5.
Table 1: Main results
(1) (2)
VARIABLES Expectations (see
methodology) Original regression: Fertility Normalized regression: Fertility GDP per capita - -0.0225*** -0.341*** (-4.27) (-4.27)
School enrollment, primary, female - 0.0104* 0.106*
(1.94) (1.94) Urbanization - 0.00105 0.0313 (0.24) (0.24) Infant mortality + 0.0276*** 0.739*** (9.62) (9.62) Time -0.320** -0.320** (-2.06) (-2.06) Treated -0.743** -0.743** (-2.17) (-2.17)
Government policy (DID-estimator) - -0.231 -0.231 (-0.67) (-0.67) Constant 1.293* 2.948*** (1.96) (1.96) Observations 276 276 R-squared 0.680 0.680
20
The effect of the One-Child policy on fertility is depicted by the ‘DID-coefficient’. After running the regression, a negative effect and a p-value of 0.505 is observed for this DID-coefficient which indicates that China’s One-Child policy had an insignificant effect on the dependent variable. On a broader level, it supports the argument that government policies are not capable of affecting fertility levels on their own. Furthermore, three out of four variables have p-values smaller than 0.1, indicating that socio-economic factors do a better job at influencing fertility. Contrary to government policies, GDP per capita (p.= 0.000), School enrollment (p.= 0.053) and Infant mortality (p.= 0.000) all significantly affect fertility in the sample.
From Table 1, the following information can be deducted regarding the effects of each socio-economic variable used in the regression. First, the results for GDP per capita suggest that a one unit increase in output per 1000 people in a country, will result in 0.0225 less birth rates. Then, a one percent increase in gross primary school enrollment among females results in a 0.0104 decrease in birth rates. Third, a one percent increase among people leaving in urban areas indicates a 0.00105 decrease in birth rates. This coefficient is rather small and points out an insignificant effect on fertility levels (p.= 0.810). Finally, a one unit increase in the mortality of infant’s results in 0.0276 more births.
Moreover, whereas the effect of GDP per capita and infant mortality is in line with the expectations made in the methodology (section 3), both school enrollment and urbanization deviate from their expected effects. Regarding education (p.=0.053), that was expected to be negatively correlated with fertility, this turns out to have a positive coefficient (0.0104). (see Table 1(1)) One possible explanation is the usage of the dataset ‘Primary school enrollment among females’. As stated in Section 3, this indicator is flawed as it lacks intel on the degree of education women end up receiving. This could also be the reason behind its weak significance (p-value=0.053).
Second, in the main results urbanization is highly insignificant (p.= 0.810). In other words, changes in urbanization have very little effect on fertility. Moreover, it is positively correlated with fertility which contradicts with the expectations. Instead of pointing towards less births, moving from a rural area to an urban one will slightly increase birth rates by 0.00105.
A possible explanation for this regards the additional opportunities that are more accessible in cities and favor having children.
Lastly, a standardized regression is used to provide a more accurate interpretation of the constant. From Table 1(2) a constant value of 2.938 can be deducted that implies that when all
21
indicators are normally distributed, there will be a fertility rate of 2.938 children per woman. As the replacement level constitutes a fertility rate of at least 2.1 births per woman, this constant may give an indication of the natural and ideal amount of births per woman, which accordingly would be 2.938. If so, then it adds on to our case that current fertility rates should be significantly higher.
Altogether, the above analysis results in the next main conclusions. The normalized regression points out that the current beyond-replacement fertility rates are indeed too low. Confirming the relevance of this paper, this means that measures should be taken to increase them to normal levels that lie around 2.938 birth rates. Then, regarding the research itself, the following can be stated. The null hypothesis is not rejected, which means that China’s One-Child policy did not have a significant contribution to the downward trend of fertility. On the other hand, it is also found that three out of the four included socio-economic factors are significant. Therefore, on a broader level, the main results suggest that it is socio-economic factors that drive the movements in fertility rates and not government policies.
Discussion
It is important to understand the implications that arise while working with the Difference-in-Difference method as they can possibly bias the results. To understand why this method was chosen, its relationship with the Fixed Effects method (FE) must be considered. In theory, the two strategies have the same basics: they both deal with fixed effects. The main difference is that opposed to FE, the DID-analysis conditions its assumptions on a group-level instead of an individual-level. And as we are comparing two groups, with the control group consisting of multiple countries, the DID-method is preferred. Moreover, the fixed-effects model is known to remove all possible effects of variables that do not vary over time, which would include the effect of the One-child policy treatment. Therefore, since government policies are the main variable of interest, the DID-method should be used.
In line with the study conducted by Bertrand, Duflo and Mullainathan (2004), a Difference-in-Difference estimation has several limitations. These often account issues regarding the standard error of the DID-estimate and refer to biases such as heteroskedasticity and serial correlation. When using this method, the observed standard errors of the ‘Government policy’ variable is likely to understate the according standard deviation. To conduct a valid study and successfully answer the research question, we need to account for these biases. This matter is addressed in Table 2 below.
22
Table 2: Validity of the model: correcting for biases
(1) (2) (3) VARIABLES Autocorrelation Correction: Fertility Robust Std Errors: Fertility 1-Log Transformation: Fertility GDP per capita 0.00434 -0.0225*** -0.0103*** (0.63) (-5.82) (-5.11)
School enrollment, primary, female 0.00306 0.0104 0.00526** (1.42) (1.49) (2.56) Urbanization -0.00291 0.00105 -0.00138 (-0.47) (0.23) (-0.83) Infant mortality 0.0579*** 0.0276*** 0.00899*** (9.15) (7.59) (8.21) Time -0.0625* -0.320** -0.117** (-1.81) (-2.19) (-1.98) Treated -0.940* -0.743*** -0.260** (-1.81) (-3.20) (-1.99)
Government policy (DID-estimator) 0.0636 -0.231 -0.135 (0.82) (-1.15) (-1.02) Constant 1.258*** 1.293 0.385 (2.76) (0.885) (1.52) Observations 253 276 276 R-squared 0.908 0.680 0.683
t-values in parentheses *** p<0.01, ** p<0.05, * p<0.1 Sample of 12 countries
Serial correlation is a common issue in Difference-in-Difference models. (see Table 2 (1)) This phenomenon implies the following. In theory, a significant effect of the One-Child policy at for example 10% level should only be observed in 10 percent of the cases. However, with autocorrelation present, there will be seemingly higher rejection rates indicating an insignificant effect. Therefore, the original results may incorrectly suggest that the One-Child policy has an insignificant effect on fertility. Several reason may induce this bias. First, the DID-estimation used has a long time series of 43 years. Second, the dependent variable ‘Fertility rates’ is likely to be positively autocorrelated. And lastly, the treatment variable, here ‘Government Policies’ does not change over time within a country. All these factors result in standard errors that are downward biased, inflating the t-statistic and by increasing the rejection rates making the p-values and final results unreliable.
To test for this issue, the Durbin Watson test is applied and gives a value of .0194234. Just as expected, a strong positive autocorrelation occurs within the dataset. To individually correct for this issue, the Prais-Winston command with a Cochran-Orcutt option can be used. (see (2) in Table 2) This increases the Durbin Watson value to 1.200956, meaning that the serial correlation issue is substantially dealt with and new results emerge. First, in line with the original results, the variable of interest indicating the effect of the One-Child policy remains insignificant (p.=0.415). However, less favorable for this research and in contrast with the main
23
results, it appears that three out of the four socio-economic factors included become insignificant. Besides Infant mortality (p.=0.000), GDP per capita (p.=0.532), School Enrollment (p.=0.156) and Urbanization (p.=0.639) are now all, highly insignificant in affecting fertility rates. Hence, this means that none of the independent variables in the model affect the dependent variable and poses a big limitation to answering the research question.
Another bias likely to arise is heteroskedasticity. This bias intervenes with the regression model its ability to predict the dependent variable Fertility consistently, across all its values. To check for heteroskedasticity, two tests were conducted: the Breusch Pagan and White’s test. Both tests have a p-value of 0.000, suggesting that the null
hypothesis is rejected and heteroskedasticity is present. The scatters in the graph depict this as their scone-like shape shows that the variance of the standard errors is not constant.
To correct for heteroskedasticity alone, two methods can be used: the 1-Log Transformation and running a robust regression. A 1-Log Transformation is applied just on the dependent variable, to decrease its variation. This then results in a decrease in the variance of the error term and deals with some of the heteroskedasticity. Second, running a robust regression gives robust standard error which tend to be more trustworthy. It gives the test more power which lowers the chance of committing type 1 errors. In the graphs below, you can see that the interval where the variance of residuals is depicted (on the vertical axis) has decreased in comparison with the original scatterplot. So, the variability of the standard errors is smaller and some of the heteroskedasticity was addressed.
After applying these methods and rerunning the regression, the following remarks can be made. First, as confirmed by the graphs, the heteroskedasticity was corrected for. Second,
24
the Government Policy variable remains insignificant for both options confirming the original results. Third, the socio-economic factors lose some of their significance. However, this phenomenon is less apparent than when correcting for autocorrelation. In this case, at most two socio-economic indicators become insignificant.
Overall, while the effect of the One-Child policy remains insignificant in all cases after addressing for biases. The results regarding the socio-economic independent factors (GDP per capita, School enrollment, Urbanization and Infant mortality) turn out less powerful in affecting the dependent variable Fertility. The latter contradicts with the original results that argue socio-economic factors to be the main drivers behind movements in fertility rates. To understand this, several limitations of this paper should be considered. Those limitations are likely the underlying reason for the contradicting results after correcting for biases in the model.
There are several limitations that should be considered. First, the methods used address the biases individually, whereas a more ideal approach would be to correct for all the biases at once. Therefore, further study on such methods should be conducted. Then, the research faces a lack of data and therefore contains only a small set of independent socio-economics factors. Finally, a smaller and more focused sample of countries similar to China may be more adequate for providing better insight in possible changes brought upon fertility by the One-Child policy. Hence, a detailed study finding such a set of countries would be advisable. Thus, to accurately assess what affects fertility the most, future research is required using a more complete dataset, additional independent variables and a more focused sample of control countries.
Conclusion
This paper aimed to analyze the significance of the relationship between China’s One-Child policy and the downward trend in fertility rates that followed. To realize this, a sample of twelve countries, including and similar to China on economic, demographic and social level, was acquired over the timespan of 1975-2016. Then, a panel data analysis was carried out using the Difference-in-Difference method. This method isolated the effect of the One-Child policy from other socio-economic factors and tried to assess which factor is the main driver behind fertility. The results indicate that the One-Child policy has an insignificant negative effect on fertility, meaning that it did not have a ‘meaningful’ contribution to the decline in fertility rates observed the years following its implementation. On a broader level, this outcome suggests that (natural) social and economic developments weigh more than government policies when it comes to influencing fertility. And therefore, using government family planning policies as a
25
method to solve the ‘graying population’ issue we are now facing will probably not be too effective on its own. So, it may be more relevant for governments to find ways of controlling the movements in socio-economic factors rather than directly controlling fertility through family planning policies. According to the empirical research conducted in this paper, such approach is likely to be successful in enhancing fertility rates and will therefore address the population aging problem.
However, it is important to note this paper encounters certain biases and limitations that should be investigated in further research. The model used appears to be subjected to autocorrelation and heteroskedasticity issues. After individually addressing for those issues, some of the socio-economic indicators lose their power and become insignificant. In other words, this means that the independent variables included have little effect on fertility levels. This forms a big limitation as it is impossible to find a way of enhancing fertility levels when these are not influenced by any factors.
Future research can improve the limitations of this paper by taking the following measures. First, different approaches should be considered that correct for all biases at the same time. Additionally, a bigger and more complete dataset of independent variables should be used. Lastly, a detailed study on countries most similar to China should be conducted. Doing this would increase the accuracy of the analysis and improve the understanding of what truly influences fertility rates. And by understanding its underlying reason, the first and most essential step in finding a solution for the issue of a declining population growth will be set.
26 Reference list
Banerjee, A. V., & Duflo, E. (2011). Poor economics: A radical rethinking of the way to fight
global poverty. New York: Public Affairs.
Aurora, V., & A. Vamvakidis, (2010). China's Economic Growth: International Spillovers.
China & World Economy, September 2011, Vol.19(5), pp.31-46
Baochang, G., Feng, W., Zhigang, G. & Erli, Z. (2007). China’s local and National Fertility Policies at the end of the Twentieth Century. Population and Development Review, March 2007, Vol.33(1), pp.129-148
Barro, R. J. (1991). Economic growth in a cross-section of countries. Quarterly Journal of
Economics 106, pp.408-443
Becker, G. (1960). An Economic Analysis of Fertility. Columbia university and National
Bureau of Economic research
Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-in-differences estimates? The quarterly journal of Economics, Vol. 119(1), pp.249-275 Bloom, D.E., Boersch-Supan, A., McGee, P. & Seike, A. (2011). Population Aging: Facts,
Challenges, and Responses. Program on the global demography of aging.
Bongaarts, J. (1994). Population policy options in the developing world. Science (N.Y.),
Vol.263(5148), pp.771-6
Bongaarts, J. (2002). The End of the Fertility Transition in the Developed World. Population
and Development Review, September 2002, Vol.28(3), pp.419-443
Bongaarts, J. (2009). Human population growth and the demographic transition. The
Population Council, Phil. Trans. R. Soc. B 364, 2985–2990
Cai, Y. (2010). China’s Below-Replacement Fertility: Government Policy or Socioeconomic Development? Population and development Review 36(3): 419–440
Cai, Y., & Wang M. (2006). Challenge Facing China’s Economic Growth in Its Aging but not Affluent Era. China & World Economy, September 2006, Vol.14(5), pp.20-31
Doepke M. (2004) Accounting for Fertility Decline During the Transition to Growth. Journal
of Economic Growth; 9(3):347–383.
Doepke, M. (2014). Gary Becker on the Quantity and Quality of Children. Northwestern
University
Easterlin, R.A. (1975) An Economic Framework for Fertility Analysis Studies in Family Planning. Population Council, Vol. 6(3) pp. 54-63
Galor, O., & Weil, D.N. (2000). Population, Technology, and Growth: From Malthusian Stagnation to the Demographic Transition and Beyond. American Economic Review, Vol.90(4), pp.806-828
27
Hilgeman, C, & Butts, C.T. (2009). Women’s employment and fertility: A welfare regime paradox. Social Science Research, Vol.38(1), pp.103-117
Hesketh, T., Lu, L., & Xing, X.W., (2005). The Effect of China’s One-Child Family Policy after 25 Years. The New England Journal of Medicine
Hopfenberg, R. (2014). An expansion of the demographic transition model: the dynamic link between agricultural productivity and population. Biodiversity, p.1-9
Li, H., Zhang, J. & Zhu, Y. (2005). The Eff ect of the One-Child Policy on Fertility in China: Identification Based on the Diff erences-in-Diff erences. Discussion Paper, Chinese
University of Hong Kong, Department of Economics
Lee, S.H., Mason, A., & Park, D. (2011). Why does population aging matter so much for Asia? Population Aging, Economic Security and Economic Growth in Asia. ERIA Discussion
Paper Series
Levin, S.M. (1938) Malthus’ Conception of the checks to population. History of Political
Economy; 12(1): 83-96
Malthus, T.R. (1798). Essay on the Principle of Population. Retrieved from:
https://www.econlib.org/library/Malthus/malPop.html?chapter_num=1#book-reader
Martine, G., Alves, J.E., & Cavenaghi, S. (2013). Urbanization and fertility decline: cashing in on structural change. IIED’s Human Settlements Group
Oded, G. (2012) The Demographic Transition: Causes and Consequences. Brown University,
Cliometrica (Berl); 6(1): 1–28
OECD (2011). Families are changing. Organisation for Economic Cooperation and Development
OECD (2014). Society at a Glance: OECD Social Indicators. Organisation for Economic
Cooperation and Development
OECD (2017). Family database. Organisation for Economic Cooperation and Development
Oztunc, H., Oo, Z.C., Serin, Z.V. (2015). Effects of Female Education on Economic Growth: A Cross Country Empirical Study. Educational Sciences: Theory and Practice, Vol.15(2), p.349-357 (Not in text)
Schultz, S. (2015), Reproducing the nation: the new German population policy and the concept of demographization. Scandinavian Journal of Social Theory, 16(3), 337-361
Settles, B.H., & Sheng, X. (2008). The one child policy and its impact on Chinese families.
Handbook of Families in Chinese Societies
Thompson, W.S. (1929) Population. American Journal of Sociology; 34(6): 959-975
Wen, P. (2014, January 10). The heartbreak behind China’s One-child policy. The Sydney
28
Yu, Z. (2008). Demographic Dynamics and Economic Take-Off - the Economic Impact of China’s Population-Control Policies. China and the World Economy for the University
29 Appendix
APPENDIX A: Main results Table 1: Main results
(1) (2)
VARIABLES Expectations (see
methodology) Original regression: Fertility Normalized regression: Fertility GDP per capita - -0.0225*** -0.341*** (-4.27) (-4.27)
School enrollment, primary, female - 0.0104* 0.106*
(1.94) (1.94) Urbanization - 0.00105 0.0313 (0.24) (0.24) Infant mortality + 0.0276*** 0.739*** (9.62) (9.62) Time -0.320** -0.320** (-2.06) (-2.06) Treated -0.743** -0.743** (-2.17) (-2.17)
Government policy (DID-estimator) - -0.231 -0.231 (-0.67) (-0.67) Constant 1.293* 2.948*** (1.96) (1.96) Observations 276 276 R-squared 0.680 0.680
Table 2: Validity of model
(1) (2) (3) VARIABLES Autocorrelation Correction: Fertility Robust Std Errors: Fertility 1-Log Transformation: Fertility GDP per capita 0.00434 -0.0225*** -0.0103*** (0.63) (-5.82) (-5.11)
School enrollment, primary, female 0.00306 0.0104 0.00526**
(1.42) (1.49) (2.56) Urbanization -0.00291 0.00105 -0.00138 (-0.47) (0.23) (-0.83) Infant mortality 0.0579*** 0.0276*** 0.00899*** (9.15) (7.59) (8.21) Time -0.0625* -0.320** -0.117** (-1.81) (-2.19) (-1.98) Treated -0.940* -0.743*** -0.260** (-1.81) (-3.20) (-1.99)
Government policy (DID-estimator) 0.0636 -0.231 -0.135 (0.82) (-1.15) (-1.02) Constant 1.258*** 1.293 0.385 (2.76) (0.885) (1.52) Observations 253 276 276 R-squared 0.908 0.680 0.683
30
APPENDIX B: How did I get to these results?
Fixed effects regression
Random effects regression
31 Difference-in-difference method
APPENDIX C: Autocorrelation test
Durbin Watson test
32
APPENDIX C: Heteroskedasticity tests
Breusch Pagan test
White’s test
Correct for heteroskedasticity: 1-log Transformation
33